Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads

[Tigerick]

Wildcat Lake (WCL) Preliminary Specs

Intel Wildcat Lake (WCL) is upcoming mobile SoC replacing ADL-N. WCL consists of 2 tiles: compute tile and PCD tile. It is true single die consists of CPU, GPU and NPU that is fabbed by 18-A process. Last time I checked, PCD tile is fabbed by TSMC N6 process. They are connected through UCIe, not D2D; a first from Intel. Expecting launching in Q2/Computex 2026. In case people don't remember AlderLake-N, I have created a table below to compare the detail specs of ADL-N and WCL. Just for fun, I am throwing LNL and upcoming Mediatek D9500 SoC.

		Intel Alder Lake - N	Intel Wildcat Lake	Intel Lunar Lake		Mediatek D9500
Launch Date		Q1-2023	Q2-2026 ?	Q3-2024		Q3-2025
Model		Intel N300	?	Core Ultra 7 268V		Dimensity 9500 5G
Dies		2	2	2		1
Node		Intel 7 + ?	Intel 18-A + TSMC N6	TSMC N3B + N6		TSMC N3P
CPU		8 E-cores	2 P-core + 4 LP E-cores	4 P-core + 4 LP E-cores		C1 1+3+4
Threads		8	6	8		8
Max Clock		3.8 GHz	?	5 GHz
L3 Cache		6 MB	?	12 MB
TDP		7 W	Fanless ?	17 W		Fanless
Memory		64-bit LPDDR5-4800	64-bit LPDDR5-6800 ?	128-bit LPDDR5X-8533		64-bit LPDDR5X-10667
Size		16 GB	?	32 GB		24 GB ?
Bandwidth			~ 55 GB/s	136 GB/s		85.6 GB/s
GPU		UHD Graphics		Arc 140V		G1 Ultra
EU / Xe		32 EU	2 Xe	8 Xe		12
Max Clock		1.25 GHz		2 GHz
NPU		NA	18 TOPS	48 TOPS		100 TOPS ?

As Hot Chips 34 starting this week, Intel will unveil technical information of upcoming Meteor Lake (MTL) and Arrow Lake (ARL), new generation platform after Raptor Lake. Both MTL and ARL represent new direction which Intel will move to multiple chiplets and combine as one SoC platform.

MTL also represents new compute tile that based on Intel 4 process which is based on EUV lithography, a first from Intel. Intel expects to ship MTL mobile SoC in 2023.

ARL will come after MTL so Intel should be shipping it in 2024, that is what Intel roadmap is telling us. ARL compute tile will be manufactured by Intel 20A process, a first from Intel to use GAA transistors called RibbonFET.

 

[Fjodor2001]

I mean they are literally desktop silicon. There is a possibility ARL-HX could be the desktop CPU tile with the mobile SOC tile and iGPU tile, but I would be pretty surprised if that was the case, I doubt it.
But I feel like this is just being pedantic lol.
I for one, think the rumor that the 8+32 desktop sku for ARL-R being canned, is very believable. And the focus on the NPU being improved on ARL-R also makes sense, I suppose they would try to improve it to the point it can hit the 40 TOPs requirement Microsoft wants?

I meant HX has different iGPU and NPU at least, plus different clocks, compared to the similar desktop CPUs. Then there is H series too. And both H and HX are mobile SKUs, and they'll have Arrow Lake Refresh.

Also, to not lose track of the background, the original claim which I commented on was "Arrow Lake refresh is just for desktop", which is simply not correct, if we are to believe the leaked Intel Mobile roadmap that I linked to previously.

 

[cannedlake240]

I've always wondered why Intel has always dedicated so much area (and presumably a decent amount of power) towards core private caches (or even cache available to the core in general) compared to the competition

Presumably to drive high clocks and as C&C says, 'insulate' the cores from slow L3. Even Apple, known for bring generous with caches among mobile soc designers, is using smaller LLC compared to Intel(4mb L2 per core vs 6mb L2 + L3 on ARL).

I suppose they would try to improve it to the point it can hit the 40 TOPs requirement Microsoft wants?

ARL igp with XMX should have enough tops for Copilot features

 

[TwistedAndy]

So 8P+32E+iGPU+Big NPU. Sounds nice. Would be lots of compute perf.

I do not think there will be any significant changes to the compute tile. It will still have an 8+16 configuration in the top desktop SKU.

The Arrow Lake Refresh will bring the updated SoC tile. It will probably have a bigger NPU and some changes to the memory controller and media engine.

Agreed. But ST perf is still a function of IPC. A 14% IPC uplift can't translate to just 3% ST perf even after factoring in all assumed limitations (like some tile penalty, some clock regression, etc). Also, that 14% is LNL. ARL should be higher. ARL with just 3% overall ST perf increase is just plain wrong imho. Looks like he has a bad sample.

Probably, it was the engineering sample with much lower clocks.

 

[mikk]

Royal core is probably still a thing, just might not be ready for NVL launch timeframe and will arrive in the generation after nvl. LNC, while an improvement over Golden cove, is still behind Z5 and M4 P cores in area, power. Caches for example, LNC has the largest amount of combined core private and LLC among general purpose high performance cores, while being at PPC parity. Royal core or not they could definitely benefit from a "ground up" core

I'm not so sure about it anymore.

Page 355 - Discussion - Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads

Page 355 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

Maybe another Ocean Cove repeat.

 

[Geddagod]

I'm not so sure about it anymore.

Page 355 - Discussion - Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads

Page 355 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

Maybe another Ocean Cove repeat.

Lmfao I went back to the tweet you linked and saw this:

@SiliconFly what changed?

Presumably to drive high clocks

Their cache density is fine tbh

ARL igp with XMX should have enough tops for Copilot features

They would presumably want to use the NPU for that though, for lower power. If they really were fine with just using an iGPU, it would make no sense that LNL got a beefy NPU, nor would it make sense why other companies are also featuring so much silicon area, as well as marketing, on NPUs.

I meant HX has different iGPU and NPU at least, plus different clocks, compared to the similar desktop CPUs.

I wouldn't be sure about that. I always assumed ARL-HX will still have the same SOC and iGPU tile as its desktop brethren.

Also, to not lose track of the background, the original claim which I commented on was "Arrow Lake refresh is just for desktop", which is simply not correct, if we are to believe the leaked Intel Mobile roadmap that I linked to previously.

👍

 

[Kepler_L2]

The "other core's" name has been leaked for a while. Panther Cove.

Pretty sure it got renamed.

 

[cannedlake240]

They would presumably want to use the NPU for that though, for lower power. If they really were fine with just using an iGPU, it would make no sense that LNL got a beefy NPU, nor would it make sense why other companies are also featuring so much silicon area, as well as marketing, on NPUs.

Process difference is huge though, N6 soc and N4P for the new gpu tile. Add to that the xmx capability and the graphics tile might be more efficient, with exception of very light workloads. Maybe 10tops isn't enough even for that, which could explain the larger NPU on the refresh

 

[SiliconFly]

-

 

[SiliconFly]

-

 

[DrMrLordX]

A 14% IPC uplift can easily translate over to just 3% ST perf. It would just mean ARL is clocking at the low 5GHz range, rather than the rumored 5.7GHz. And the final version could end up being 5.7GHz, since he even says he has an ES2, but the final ST perf uplift is still likely to be pretty low, as in like 10% perhaps, at best.

There's still that leaked Intel slide from last year showing +5% ST and +20% MT over Raptor Lake at the same core count and PL value. Take out HT from the Arrow Lake P-cores and then use the perf numbers from the 14900k/ks before Intel implemented their stability fixes for UE5 (and other software) and that could still bear out. By the time Arrow Lake comes out it will look like a big uplift over Raptor Lake because of the new microcode updates and performance profiles gimping Raptor Lake performance, but that's not the only comparison that will matter . . .

 

[Hulk]

Could AMD or Intel develop a compiler specifically designed to work optimally with their specific CPU's? If so could such a compiler be used to compile compute heavy applications so that the OS would "pick" the correct version of the software upon installation?

The reason I ask is that it seems we are approaching a point where to optimize hardware, software might have to be compiled with the hardware architecture taken into account.

 

[poke01]

one thing Intel is really good is supporting a wide array of encoders and decoders, no AMD or Nvidia GPU supports 10 bit 422 HEVC encode/decode.

So there are reasons for creators to get arrow lake CPUs over AMD.

https://twitter.com/x/status/1809715549166862487

 

[DavidC1]

There's still that leaked Intel slide from last year showing +5% ST and +20% MT over Raptor Lake at the same core count and PL value. Take out HT from the Arrow Lake P-cores and then use the perf numbers from the 14900k/ks before Intel implemented their stability fixes for UE5 (and other software) and that could still bear out.

So it was showing those gains against 13900K.

Intel’s internal performance projection for Raptor Lake S Refresh and Arrow Lake S - How fast the CPU and iGP are expected to be | Exclusive | igor´sLAB

The Raptor Lake refresh is expected to arrive this year as the 14th CPU generation for the desktop. But the real successor, Arrow Lake-S for desktop, is also increasingly materializing along with the…

www.igorslab.de

RPL: 8x1.3 + 8 = 18.4
ARL Int: 8x1.14 + 8x1.3 = 19.52(6%)
ARL FP: 8x1.14 + 8x1.6 = 21.92(19%)

Those calculated gains seem to be not far off from Igorlabs leak. We really should not expect LNC in ARL to be higher than 14%. Nothing beside hopeful guesses seem to point ARL to be a monster.

I think when they are saying the design is now more flexible thus faster iterations can happen they are talking about primarily regarding server vs client. Remember the 8-wide decode for Granite Rapids?

 

[Geddagod]

Process difference is huge though, N6 soc and N4P for the new gpu tile. Add to that the xmx capability and the graphics tile might be more efficient, with exception of very light workloads. Maybe 10tops isn't enough even for that, which could explain the larger NPU on the refresh

Well, you answered your own question at the end there...

Pretty sure it got renamed.

I'm sure Panther Cove being in Nova Lake rather than Panther Lake wasn't confusing at all

 

[DavidC1]

And one more thing...

This one is from the Intel P core team engineer. Lion Cove version that goes into Lunar Lake is different in several aspects compared to the version that goes into Arrow Lake later this year.

They may not even be close! That >20% uplift is still very much a possibility.

Yea you are likely way overestimating the impact. I made such mistakes before when they talked about Sandy Bridge's ring bus allowing much faster iterations. They had more versions for sure, but it didn't come faster. The rate of increase in complexity is so steep that they need such drastic design approach changes just so it doesn't take longer and longer to come out.

This time is different folks. It's not just Intel starting to lose steam this time. It's both Intel and AMD. AMD followed Intel to the crazy clock focus and they are following Intel with the disappointing 16% gain. They are going to need radical overhaul of designs/design teams to counterract ARM.

Looks like Intel has a path forward with the Austin team.

 

[poke01]

one thing Intel is really good is supporting a wide array of encoders and decoders, no AMD or Nvidia GPU supports 10 bit 422 HEVC encode/decode.

So there are reasons for creators to get arrow lake CPUs over AMD.

https://twitter.com/x/status/1809715549166862487

Coming back to this, Intel and Apple media engines are the only ones that support 422 10 bit HEVC.

 

[FlameTail]

Coming back to this, Intel and Apple media engines are the only ones that support 422 10 bit HEVC.

And what's the significance of 422 10-bit HEVC?

 

[poke01]

And what's the significance of 422 10-bit HEVC?

Its for videographers. Canon, Fujifilm cameras have the ability to shoot this

You got a RTX 4090 and a 7950x and still you won't be able to smoothly playback 422 without the media engine supporting these codecs.

 

[Hitman928]

Its for videographers. Canon, Fujifilm cameras have the ability to shoot this

You got a RTX 4090 and a 7950x and still you won't be able to smoothly playback 422 without the media engine supporting these codecs.

They can do 444 and 420 playback, just not 422. I don’t know how big a deal that is for creators outside of some professional cameras that use 422 (can they use 444? No idea).

 

[Hulk]

422/444 are editing formats, 420 is a delivery format.
I would characterize 422 as semi-pro and 444 as pro.
422 holds up much better in post than 420.

 

[Det0x]

Not sure if this already have been posted

https://twitter.com/x/status/1809987963582890343

 

[poke01]

Not sure if this already have been posted
View attachment 102543

https://twitter.com/x/status/1809987963582890343

This is not that bad. Intel first needs to fix their efficiency and I rather they do that than go all out in performance.

 

[Kepler_L2]

This is not that bad. Intel first needs to fix their efficiency and I rather they do that than go all out in performance.

Who says power consumption went down?

 

[poke01]

Who says power consumption went down?

there has to be some performance per watt improvement right?

I mean they’re jumping from Intel 7 to N3B.

Edit: Those numbers are at 250w, crazy. Wtf, is Intel doing?

 

[KompuKare]

there has to be some performance per watt improvement right?

I mean they’re jumping from Intel 7 to N3B

Chiplets are not free of an energy penalty though. How much? That's anyone's guess for now.